Distinctive epigenomes characterize glioma stem cells and their response to differentiation cues
Keith D. Robertson
Primary patient GBM tissue was transplanted into mice as described previously to create patient derived xenografts (PDX) . The stem-cell population was isolated from 22 different PDX tumors (Additional file 1: Tables S1 and S2), whereas neural progenitor lines, a hypothetical origin of GSCs, were isolated from fetal brain (NSC23, NSC27, and NSC30). Both cell types formed neurospheres in non-adherent culture conditions but attached onto a laminin/ fibronectin-coated surface (Fig.
1a left, Fig.
1b left). Immunofluorescence (IF) for stem markers NESTIN and SOX-2 were detected in NSCs (Fig.
1b). The stemness of GSCs was assessed by staining for NESTIN, SOX-2, and CD44, and lineage markers GFAP (astrocyte), TUBB3 (neuron), and GALC (oligodendrocyte) (Fig.
1a, Additional file 1: Table S1, top). Twelve GSC lines containing a high proportion of stem marker-positive and lineage marker-negative cells were used for further molecular analyses (Fig.
1c, Additional file 1: Table S2). Indeed, consistent with these stem and lineage marker expression patterns, genome-wide 5mC patterns based on 450 k array analysis (run on 22 GSC lines and eight glioma cell lines) also largely segregates these 12 GSC lines into a distinct group (Additional file 2: Figure S1A).Expression of the DNA modification machinery in GSCs is variable and distinct from that in NSCs. TET1 transcription is overall downregulated, whereas TET2 and TDG are upregulated in the majority of GSCs; TET3 is the most variably expressed (Fig.
1d). Global quantification of DNA marks using mass spectrometry (MS)  show that NSCs overall possess higher amounts of 5mC and 5hmC, but lower levels of 5fC, compared to GSCs (Additional file 2: Figure S1B). Their levels were further compared to those of established serum cultured glioma cell lines T98G and A172, and normal brain tissue. Notably, the cancer cell lines exhibit lower levels of 5mC compared to normal tissue (Additional file 2: Figure S1C top), but much reduced levels of 5hmC and 5fC (Additional file 2: Figure S1C middle and bottom). 5caC levels are not shown because our LC-MS/MS method failed to capture the 5caC-containing fractions from all samples analyzed. In keeping with these findings, PDX tissue derived from GBM6, GBM64, and GBM84 exhibit reduced immunohistochemical (IHC) staining intensity in the nucleus for both 5mC and 5hmC, relative to surrounding normal tissue (Fig.
1e). We correlated transcription level of the TETs to the level of each DNA mark (Additional file 2: Figure S1D) and observed that TET2 and TDG expression are significantly negatively correlated with 5mC level, while TET2 is positively correlated with 5fC (Fig.
1f, Additional file 2: Figure S1D), suggesting that TET2 expression contributes to the increased level of 5fC in GSCs.Tumor tissue from a well-characterized panel of GBM patient-derived xenografts (PDX) developed as part of the Mayo Clinic Brain SPORE, was mechanically dissociated and selected for the stem-cell population in stem-cell culture medium (ThermoFisher, StemPro NSC SFM A1050901) containing L-glutamine and penicillin-streptomycin using laminin (#L2020) coated flasks to reduce culture condition-introduced heterogeneity [15, 58, 59]. Fetal brain-derived neural stem cells, NSC23, NSC27, and NSC30, were obtained from Dr. Philip H. Schwartz at the Children’s Hospital at Orange County (CHOC) . Neural stem cells were expanded on Matrigel (Corning 356,230) or fibronectin-coated 6-well plates in stem-cell growth medium (GM) containing DMEM/F-12 plus Glutamax (Life Technologies 10,565–018), BIT9500 serum substitute (1×, Stem Cell Technologies 09500), heparin (20 μg/mL Sigma H3149), primocin (1×, InvivoGen ant-pm-1), human bFGF (20 ng/mL, Stemgent 03–0002), and human EGF (20 ng/mL, R&D Biosystems 236-EG) . Low passage cultures of cells (< 5 and typically < 3 passages) were used for all experiments and deep sequencing.We kindly thank Yue Wang for developing phython programs for data analysis, Ann Tuma and Alissa Caron for GBM tissue samples, and Leonard Collins for conducting LC/MS-MS.All experimental animal procedures described in this work were reviewed and approved by the Mayo Clinic Institutional Animal Care and Use Committee (#A56814–14, “Creation, maintenance and characterization of brain tumor xenograft panel”). All research involving human tumor material was performed in accordance with the Declaration of Helsinki. Patient brain tumor samples for mouse xenograft generation were obtained with informed consent under a Mayo Clinic Institutional Review Board approved protocol (#07–007623, “Primary xenograft model for studying brain tumor biology and therapy”).Electronic supplementary materialThe online version of this article (10.1186/s13059-018-1420-6) contains supplementary material, which is available to authorized users.